Expert Systems in Accounting in a Personal Computer Environment
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Expert Systems in Accounting in a
Personal Computer Environment
DANIEL E. O'LEARY*
Introduction
A recent survey by the AICPA [Frotman and Wixson, 19851
found that the use of the personal computer is rapidly growing.
The software commonly available for the personal computer in
cludes word processing, database programs, and spreadsheet
programs.
Recently, expert systems and artificial intelligence tools have
become available for the personal computer. Since accountants
have just begun to use artificial intelligence and develop expert
systems, there has been only limited analysis of the implications of
these new tools for the profession. Accordingly, the purpose of this
paper is to analyze the development of expert systems and the use
of artificial intelligence technology in accounting in a personal
computer environment. l '
This article will discuss the following subjects:
• review of artificial intelligence and expert systems,
• analysis of artificial intelligence and expert systems in
accounting,
• the impact of the personal computer environment on
accountants,
• the impact of expert systems in accounting,
• the use of artificial intelligence and expert system shells for
expert system development on a personal computer, and
• summary of expert systems developed for accountants.
.. Daniel E. O'Leary, Ph.D., is an Assistant Professor of Accounting at The University of
Southern California.
107108 GEORGIA JOURNAL OF ACCOUNTING, SPRING 1986
Artificial Intelligence and Expert Systems
Rich [1983, p. 1] defined artificial intelligence (AI) as "...the
study of how to make computers do things which, for the moment,
people are better." Barr and Feigenbaum [1981, p. 1] have defined
AI as ". . . the part of computer science concerned with designing
intelligent computer systems, that is, systems that exhibit the
characteristics associated with intelligence in human behavior...
." These definitions indicate that AI is the study of developing
computer systems to perform tasks and do analysis that humans
currently use knowledge and reasoning to carry-out.
The domains of AI include knowledge representation in com
puters, natural language use with computers, learning by com
puters, and other topics. Good surveys are included in Barr and
Feigenbaum (1981) and Rich (1983).
Knowledge-based expert systems (ES) are also a branch of AI.
ES's perform tasks normally done by knowledgeable human ex
perts [Rich, 1983]. Accordingly, ES's are developed by program
ming the computer to make decisions using the processes and
knowledge as the expert.
Typically, ES's perform intellectually demanding tasks, rather
than mechanical tasks. In addition, ES's usually have the ability to
explain their reasoning [Barr and Feigenbaum, 1981].
Structurally, ES's usually have four major components:
Database, Knowledge Base, Inference Engine, and User Interface.
The database includes the data used by the expert system. This is
normally the same data that a human expert uses to solve the
problem. The data may be a part of the program, it may be a part
of the database, or it may be elicited from the user.
The knowledge base provides the set of knowledge that the ex
pert indicates is used to process the data. The knowledge base con
tains the knowledge that the ES uses to process the data. Typi
cally, this is the domain-specific knowledge that an expert would
use to solve the problem. Knowledge can be represented in a num
ber of ways. Two of the primary methods are rule-based and
frame-based knowledge representation. Rule-based knowledge rep
resentation generally takes the form of "if ... (condition) then
... (consequence/subgoal/goal)." The rules mayor may not in
clude a numeric level of confidence or probability of occurrence.
Frame-based knowledge representation uses a "frame" to capture
the characteristics associated with a given entity. CharacteristicsO'LEARY 109
define the knowledge that is of interest to the entity.
The inference engine provides the basis to use the knowledge
base to process the database. In a rule-based system, the inference
engine normally uses either a forward or backward chaining ap
proach. Forward chaining reasons toward a goal. Backward chain
ing reasons backward from the goal to determine if or how the goal
can be accomplished.
The user interface defines the relationship between the user
and the system. Generally, the interface is user friendly, particu
larly in those situations where the data is generated from the user.
Expertise and Decision Characteristics
Researchers have identified some characteristics about success
ful expert systems. First, the expertise that is being modeled
should be in short supply [Fox, 1984], an expensive resource or not
readily available at a particular geographic location. Second, a dif
ference should exist between the decisions of an expert and an am
ateur [McDermott, 1984]. Otherwise, there is no need for such an
expert. Third, the knowledge base and the inference engine must
not be easily acquired. Otherwise, the user could simple develop
the knowledge base and the inference engine and operate indepen
dently of the expert. Fourth, Fox (1984) has suggested that the de
cision should require short reaction time or else the expertise may
be developed by the user. Fifth, McDermott (1984) adds that the
decision should be a high value decision. These conditions ensure
that there is a high cost-benefit contribution of the ES.
Purpose of System
AI and ES have been used to develop programs that perform in
an educational mode, an advisory mode, and a replacement mode.
Educational AIlES are being used for modeling educational
functions that previously would not have been placed in a com
puter model. STEAMER [Williams et al., 1981] is an example of a
simulation program that uses concepts from AI to serve as a tutor,
training students in the principles of propulsion engineering.
Advisory Most existing expert systems are designed to func
tion in an advisory manner. These systems make a recommenda
tion and a human expert reviews the decision and the logic behind110 GEORGIA JOURNAL OF ACCOUNTING. SPRING 1986
the decision before the decision is executed.
Replacement There are some systems designed, however, to
replace the decision maker. Glover et ale (1984) designed a system
that they suggested should be called a "managerial robot" because
it was designed to replace the decision maker. The system was
designed to schedule employees in an environment of weekly
fluctuations.
Although the system was designed to replace the manager, it
does not have to be implemented in that manner. The system can
be implemented to advise the user. (It is interesting to note that
this system may be used for scheduling auditors.)
Implications of Expert Systems in Accounting
The expertise and decision characteristics of a successful expert
systems suggest that accountants function in an environment
where ES's can aid the user.
Expertise and Decision Characteristics
Much accounting expertise is in short supply and not always
easily accessible. In order to mitigate the limited supply problem
of expertise many of the larger public accounting firms have cen
trally located expertise (e.g., industry and function). Demand for
an accountant's expertise can arise at the client's office where the
desired expert may not be available, or the expert may receive
multiple calls from the field simultaneously and thus not be availa
ble for each of the requests.
The knowledge of the accountant is not easily obtained. In or
der to pass the CPA exam a large body of knowledge must be as
similated. Becoming a partner in a public accounting firm requires
additional knowledge. Accounting expertise is also expensive. Cur
rent billing rates of CPA's attest to the high cost of expertise.
Accounting decisions are frequently high value decisions. For
example, the litigatory nature of some audit decisions associates a
large potential cost with those decisions.
Uses of Expert Systems
Accountants can use ES's for education, and advising. Account
ing firms do substantial amounts of training of their personnel. ItO'LEARY 111
may be that ES's could be used in training accountants. As noted
below, there have been a number of prototype expert systems de·
veloped for advising accountants on a variety of issues.
Personal Computer Environment
The personal computer (PC) environment arises from individ
ual use of the personal computer. The personal computer environ
ment is differentiated from either minicomputers or mainframe
computer environments in at least four ways. First, the broad dis·
semination of PC's allows the user to use their own PC at their
home office location and another's compatible PC at their work lo~
cation (e.g., at the client's). Even if the alternate work location
does not have a PC, portable PC's are available that allow the user
to take the computer to the problem. This allows the user to have
the same software at both locations and, accordingly, implement
the power of the computer at both locations. Second, information
storage is private (as opposed to publicly available) and under con·
trol of the user. The physical control of a diskette, for example,
means that information and knowledge can be used at arbitrary
locations in the privacy of the user's office. In addition, the user
can physically protect the data from unauthorized use by control·
ling the diskettes. Third, the user has direct personal control over
all stages of the activity [Keen and Hackathorn, 1979J ensuring
control over quality and a necessary understanding of the use of
the programs. In addition, the user can choose the tasks and allo·
cate the resources for which software is developed or purchased.
Accordingly, this can lead to the development or purchase of
software in a timely manner. A computer department can no
longer be used as a reference or an excuse. Fourth, the software
and programs that have been developed for the PC are normally
very user friendly and inexpensive when compared to mainframe
software and programs. User friendliness allows the use of the sys
tem for training, limits the time required for training, increases the
scope of use, and allows the user to be familiar with a larger base
of software. Inexpensive software leads to the user being familiar
with a broader base of programs because of affordability.112 GEORGIA JOURNAL OF ACCOUNTING, SPRING 1986
ES in Accounting in a PC Environment
Four primary implications for accountants derive from the use
of ES's in the PC environment. First, the dissemination of PC's
and the portability of PC's allow the accountant (e.g., auditor or
consultant) to have computer capabilities at virtually all locations
(e.g., at the client's). Thus, the accountant can bring ES capabili
ties on-site. To the extent that AI or expert knowledge can be cap
tured in a computer program, that intelligence and expertise can
be used in virtually all locations. Consequently, high value or high
CQ8t decisions can be made in consultation with an expert system
at virtually any location. Second, the storage of private informa
tion allows the accountant to carry the expertise of a number of
tasks to the site of the activity and still ensure privacy of the ex
pertise. This leads to mobile and available expertise. Privacy also
allows the use of arbitrary AI-based or ES training tools. Third,
the personal control of the activity allows the accountant to choose
the areas of ES application to meet the user's needs and suggests
that the accountant may develop the applications. Fourth, the user
friendly and the inexpensive aspects of the PC environment make
the AIlES tools usable and economically feasible.
AIlES Capabilities on PC's
Artificial intelligence techniques are implimented and expert
systems are developed using three primary approaches: procedural
languages, artificial intelligence languages, and expert system
"shells." Only recently have the last two methods been available
for use on a PC.
Procedural Languages
Procedural languages, such as BASIC, allow the user to define a
sequenced set of operations to solve a specific problem. Some ex
pert system shells and some expert systems have been developed
using procedural languages. Fortran and Pascal are the two most
frequently used procedural languages in the development of AI ap
plications or expert systems. Both Fortran and Pascal are available
on the PC [e.g., Borland, 1985].O'LEARY 113
Artificial Intelligence Languages
Two primary artificial intelligence languages are in use: Prolog
[Clocksin and Mellish, 1984] and LISP [Whinston and Hom,
1984]. Prolog has been chosen by the Japanese for their fifth gen
eration computer project. Whereas, the primary AI programming
language applications in the United States have used LISP. Both
these languages are used on personal computers [e.g., Clark and
McCabe, 1984 and Steele, 1984, respectively].
AI languages differ from procedural languages in three primary
ways. First, in contrast to other computer languages that are
designed to process specific numeric information, an AI language is
designed to process language-based and abstract symbol informa
tion. Second, the procedural languages are dependent on the order
of the statements, whereas LISP does not have the same proce
dural constraints [Sheil, 198]. Third, unlike procedural languages,
Prolog is a natural and easy-to-analyze language.
In some cases the designer may use both AI and other lan
guages. Some versions of Prolog and LISP allow the user to access
procedural languages. The advantages and limitations of Prolog
and LISP have been summarized by Williamson (1984).
1. Prolog's main advantage is its natural way of expressing
ideas. The coded knowledge can be analyzed by the non-pro
gramming expert for its accuracy and completeness. LISP is
more difficult to follow.
2. The order of the knowledge is important in Prolog but not in
LISP. As a result, LISP programs are easier to maintain.
3. Prolog's development times are usually shorter than LISP's.
Thus, it may be less expensive to develop a Prolog
application.
Expert System Shells
Expert system shells simplify the development of an expert sys
tem by providing many user friendly features [e.g., Turpin, 1985].
The inference engine can be specified and does not need to be de
veloped. The knowledge base usually is easy to program (not nec
essarily easy to elicit from the expert). The shells also may allow
the user to access existing databases such as dBase III and other114 GEORGIA JOURNAL OF ACCOUNTING, SPRING 1986
AI languages. Some shells (e.g., M.1) are designed for program
mers, while others (e.g., Insight) are easier to use [Williamson,
1985]. The cost of the shells ranges from around $100 (e.g., Insight)
to over $10,000 (e.g., M.1). A number of these shells are listed in
the appendix.
Which Approach Should Be Used?
The most basic (and generally the most difficult) approach to
the development of an expert system or other artificial intelligence
application is to program the system using a procedural program
ming language. The difficulty arises because these languages were
not developed primarily to process symbolic information and do
not have the unique features that specialization brings. A second
approach is the AI language. However, these languages do not have
the user friendly features of the shells. Accordingly, to the extent
that a shell can be used to develop the system, that option may be
executed. In either case, the AIlES system developer should choose
that approach that best meets the developer's needs in a cost effec
tive manner.
Accounting Applications of AI/ES2
There have been few applications of AI and ES in accounting.
However, those applications that have been developed provide evi
dence of the use of a range of computer languages and shells in the
development of AI and ES in accounting. These systems also indi
cate a broad potential scope of applications. Each of these applica
tions could be useful to the accountant both in the accountant's
and the client's offices. This paper treats the applications as falling
into one of two categories: systems in use or prototype systems.
Most of the systems to date are prototype systems. Currently, ES's
in particular and AI in general are not widely used.
Prototype Applications
The primary analysis of expert systems in accounting to date,
has been preliminary in nature. The previous research has concen
trated on the feasibility of developing accounting-based expertO'LEARY 115
systems.
Taxadvisor Taxadvisor is a prototype expert system developed
by Michaelsen (1982, 1984) for estate planning. Taxadvisor was de
veloped using the shell Emycin [Buchanon and Shortliffe, 1984].
Auditor Auditor is a prototype expert system developed by
Dungan (1983) and Dungan and Chandler (1983) for the evaluation
of the adequacy of the allowance for bad debts decision. Auditor
was developed using the shell AL/X [Reiter, 1980].
EDP Auditor EDP Auditor is a prototype expert system devel
oped by Hansen and Messier (1982, 1985). EDP Auditor is an ex
pert system for auditing computer-based accounting systems. EDP
Auditor was developed using a shell AL/X [Reiter, 1980].
ICE ICE is a prototype expert system developed by Kelly (1985)
for internal control evaluation. ICE was developed using a dialect
of LISP.
TICOM TICOM is a prototype, computer-based tool based on
some AI concepts developed by Bailey et al. (1984a and 1984b).
TICOM is used to aid the auditor in the analysis of the internal
control system. TICOM was developed using Pascal. The develop
ers of that system suggest that it be interfaced with an expert
system.
Conclusion
This paper suggests that using AI techniques and ES's in ac
counting in a PC environment capitalizes on the strengths of the
PC and matches the characteristics of accounting expertise and
knowledge. The PC allows the user to use AIlES training capabili
ties. It makes the location of the training independent of the ex
pert and increases the scope of potential training activities. The
PC also allows the user to model a variety of advisory expertise
where the expertise is expensive and not accessible. The PC envi
ronment allows the user to take the expertise to off-site
environments.
APPENDIX
The purpose of this appendix is to provide a list of some of the
available shells for personal computers. This list derives from116 GEORGIA JOURNAL OF ACCOUNTING, SPRING 1986
Miller (1984), Harmon and King (1985), Williamson (1985) and di
rect contact with some of the vendors. The addresses of these ven
dors are in those references. There new packages being released all
the time so this is not a complete list.
Shell Company
AL/X University of Edinburgh
ES/P Advisor Expert Stems International
Expert Ease Expert Software International
Expert Edge Human Edge Software
EXSYS Essays Inc.
Insight Level Five Research
Insight2 Level Five Research
KAS SRI
KDA KDS Corp.
KES Software A & E
KES II Software A & E
M.1 Teknowledge
Pesonal Consultant Texas Instruments
TIMM General Research
Series-PC SRI
RuleMaster Radian Corp.
XSYS California Intelligence
NOTES
I This analysis is oriented toward personal computers such as the IBM PC, AT&T PC,
IT&T PC, COMPAQ, and the Apple PC's, but it does not include the so called "LISP"
machines such as the XEROX 1100.
• The expert systems evaluated in this article were found by examining four sources,
including Peat, Marwick, Mitchell & Co.'s Research Opportunities in Auditing (1985).
Miller's 1984 Inventory of Expert Systems (1984), Ph.D. Dissertations through calendar
year 1984 (because of the lag with their being listed in University Microfilms) and from the
author's awareness of particular papers and presentations.
References
Bailey, A. D., Duke, G. L., Gerlach, J .• Ko, C.• Meservy, R. D., and Whinston, A. B.,
"TICOM and the Analysis of Internal Controls," Working Paper, University of Minne
sota, 1984.
Bailey, A. D., Duke, G. L.• Gerlach, J., Ko, C., Meservy, R. D., and Whinston, A. B.• "Com
puter Assisted Evaluation of Internal Controls: TICOM III," Working Paper, University
of Minnesota, 1984.
Barr A. and Feigenbaum, E. A., The Handbook of ArtijicialIntelligence Volume I, HeurisO'LEARY 117 tech Press, Stanford, Ca. and William Kaufmann, Los Angeles, Ca., 1981. Buchanan, B. G. and Shortliffe, E. h., Rule-Based Expert Systems, Addison-Wesley, Read ing, Massachusetts, 1984. Clark, K L. and McGabe, F. G., Micro Prolog: Programming in Logic, Prentice-Hall, Engle· wood Cliffs, New Jersey, 1984. Clocksin, W. F. and Mellish, C. S., Programming in Prolog, Springer- Verlag, New York, 1984. Dungan, C. and Chandler, J. S., "Analysis of Expert Judgement Through an Expert Sys tem," Working Paper 982, College of Commerce and Business Administration, 'Univer sity of Illinois, November, 1983. Fox, M., "Artificial Intelligence in Manufacturing," paper presented at the CPMS Seminar on Expert Systems, December, 1984. Frotman, A. and Waxson, J. A., "Micros in Accounting," Journal of Accountancy, October, 1985, pp. 136-150. Glover, F., McMillan, C. and Glover, R, "A Heuristic Programming Approach to the Em ployee Scheduling Problem and Some Thoughts on 'Managerial Robots'," Journal of Operations Management, Vol. 4, No.2, 1984, pp. 113·128. Hansen, J. V. and Messier, W. F., Expert Systems For Decision Support in EDP Auditing," Internationals Journal of Computer and Information Sciences, Vol. 11, No.5, 1982. Hansem, J. V. and Messier, W. F., "A Knowledge-Based Expert System for Auditing Ad vanced Computer Systems," Working Paper, January, 1985. Harmon, P. and King, D., Expert Systems, John Wiley & Sons, New York, 1985. Hayes-Roth, F., Waterman, D. A. and Lenat, D. B., Building Expert Systems, Addison Wesley, Reading, MA, 1983. Keen, P. G. W. and Hackathorn, R. D., "Decision Support Systems and Personal Computer ing," Department of Decision Sciences, University of Pittsburg, 1985. Kelly, K P., Expert Problem Solving System For the Audit Planning Process, Unpublished Ph. D. Dissertation, University of Pittsburg, 1985. McDermott, J., "Background, Theory and Implementation of Expert Systems II," paper presented at the CPMS Seminar on Expert Systems, December, 1984. Michaelsen, R., "An Expert System for Federal Tax Planning," Expert Systems, Vol. 1, No. 2, pp. 149-167, 1984. Miller, R. K, The 1984 Inventory of Expert Systems, SEAl Institute, Madison, GA, and Technical Insights, Fort Lee, New Jersey, 1984. Peat, Marwick, Mitchell & Co., Peat Marwick Foundation Research Opportunities in Au diting, Interim Report. 1985. Retter, J., AL/X: An Expert System using Plausible Reasoning, Intelligent Terminals, 1980. Reitman, W. (ed), Artificial Intelligence Applications for Business, Ablex Publishing, Nor- wood, New Jersey, 1984. Rich, E., Artificial Intelligence, Mcgraw-Hill, New York, 1983. Sheil, B., "The Artificial Intelligence Tool Box," in Reitman (1984), pp. 287-295. Tello, E., "Raw Power for Problem Solving," PC Magazine, April 16, 1985. Teschler, L., "Stripping the Mystery from Expert Systems," Machine Design, April 25, 1985. Turpin, W., Personal Consultant Plus: Expert System Development Tools, Technical Re port, Texas Instruments, 1985. Steele, G. L., Common LISP (Reference Manual), Digital Equipment, 1984. Whinston, P. H. and Horn, B. K P., LISP, Addison-Wesley, Reading Massachusetts, 1984. Williams, M., Hollan, J. and Stevens, A., "An Overview of STEAMER: An Advanced com puter-Assisted Instruction System for Propulsion Engineering," Behavior Research Methods and Instrumentation, VoL 13, No.2, 1981, pp. 85-90. Williamson, M., "Knowledge-Based Systems," PC Week, July 9, 1985.
Volume Seven Spring 1986
MICROCOMPUTERS: THE EMERGING REVOLUTION IN ACCOUNTING
Impact of Computerization on the Profession
Stanley D. Halper .................... . 1
Computer Technology and Training: Yesterday,
Today. and Tomorrow
Gerald F. Hunter and Mark A. Poplis ...... . ................... 15
l each Spring Office Automation in Exxon Company, U.S.A.:
University of A Tool for Change
Michael C. Wilser. . . . . . . . . . . . . . . . . . . .. . ........... . . ......... 41
e Journal in
ge. Although Microcomputers in Management Information Consulting
Jack Wilson and Glover Ferguson . . . . . . . . . . . . . . ...................... , 61
;0 the School
The Use of Computers in a CPA Practice:
lUtomatically
A Private Companies Practice Section Perspective
Itributions of Thomas L. Jollay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 75
nated for use The Use of Microcomputers in the Banking
Industry: Trust Company of Georgia
.e GEORGIA Patricia R. Grimes ................ . .. ............................. 89
,n request as Expert Systems in Accounting in a
Personal Computer Environment
Daniel E. O'Leary .... ..................................... . ......... . 107
lAL OF AC
A Microcomputer is not Just a Small Computer:
, of Georgia; An Internal Control Perspective
John S. Chandler . . . . . .. ............ . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 119
Departments:
rights reserved. Microcomputer Audit Techniques
J. Porter Bellew. Jr.......................... . . ........................ 131You can also read
